You are on page 1of 17

28

Transient heat conduction


The steady-state solutions studied in the previous section usually are achieved physically
when the system has evolved for a sufficiently long period of time from a specific initial state.
Here, we are concerned with the temporal evolution of the system. Considering a constant
thermal conductivity, the heat conduction equation (45) can be expressed as follows,

∂T E
= α∇ 2 T + r (131)
∂t ρC p
where
k
α= (132)
ρC p

is the thermal diffusivity.


In general, equation (131) will be a partial differential equation having as independent
variables time and spatial coordinates. Since the equation is first order with respect to time, an
initial condition is required. Typically, this condition is the specification of the temperature
profile at t=0; in rectangular coordinates we will have

T=T0(x,y,z), t=0 (133)

where T0 is known.
For the study of transient conduction, we will analyze the common practical problem of a
body surrounded by a fluid, so that the heat transfer between the body and the surrounding fluid
can be characterized by a convective heat transfer coefficient. For this case, certain
simplifications can be made, depending on the relative importance of internal conduction in the
body with respect to external convection. As we saw in the previous section, this comparison can
be based on the Biot number, which we will define generally as

hL c
Bi = (134)
k

where Lc is a characteristic length representative of the conduction process.


For example, consider a large flat plate of thickness b with uniform initial temperature T0 that
is suddenly exposed on both sides to a fluid at T∞<T0. Heat transfer from the wall to the fluid will
induce a temperature profile. In this case, we can use as characteristic length in the definition of
the Biot number Lc=b. The shape of the temperature profiles that will evolve with time will
depend on the relative magnitude of the Biot number (Figure 11):
(a) For Bi«1, conduction inside the wall will be fast compared to external convection. This
means that the temperature in the wall will be approximately uniform at all times (a quantitative
analysis of this case will be similar to that performed in the context of Figure 9).
(b) For intermediate Biot numbers, appreciable profiles will exist both inside the wall and in
the fluid.
(c) For Bi»1, convection is fast compared to conduction. As a consequence, the fluid
temperature will be approximately uniform. This means that the surface temperature of the wall
reaches T∞ a short time after the process starts, and remains at that value for all times.
29

Figure 11. Evolution of temperature profiles in a flat wall in contact with a fluid at T∞.
30

Note that, unless Bi«1, the temperature in the wall will be a function of position and time.
However, for Bi«1, the wall temperature can be considered uniform and it will be only a function
of time (as an approximation). In this case we say that the wall has negligible internal resistance
to heat transfer and the heat conduction equation is considerably simplified.

Transient heat conduction with negligible internal resistance


If internal resistance to heat transfer is negligible, the temperature inside a body can be
considered uniform and, therefore, the shape of the body does not affect the heat transfer process
or the evolution of its temperature. We will perform our analysis considering a body with
arbitrary shape. An important consideration is the choice of characteristic length to use in the
definition of the Biot number. Since we will require Bi«1, we will select the largest characteristic
dimension associated with the body (Figure 12). The heat flux at the surface of the body is
determined at each point by an energy balance stating that the normal component of the
conductive heat flux vector (qn) must equal the heat transferred by convection to the surrounding
fluid.

Figure 12. Heat transfer from a body with negligible internal resistance, as dictated by
Bi=hLc/k«1, where Lc is chosen as the largest dimension associated with the body. At each point
on the surface, the heat conducted in the direction normal to the surface must equal the heat
transferred by convection into the fluid.

To analyze this case, we start with the heat conduction equation (45) expressed in terms of the
heat flux vector:
∂T
ρC p = −∇ ⋅ q + E r (135)
∂t

We integrate this equation over the whole volume of the body, V:

∂T
∫ ρC p ∂t dV = − ∫ ∇ ⋅ qdV + ∫ E r dV (136)
V V V

Since the temperature is uniform, all factors inside the integrand on the left-hand side of this
equation are uniform and, therefore, we get:
31

∂T dT
∫ ρC p ∂t
dV = ρC p
dt
V (137)
V

where the partial derivative has been converted to a total derivative since T=T(t). We will
consider also that Er is uniform within the body, so that

∫ E r dV = E r V (138)
V

Finally, the heat flux term can be converted into an area integral using the divergence theorem:

∫ ∇ ⋅ qdV = ∫ q n dA = ∫ h (T − T∞ )dA (139)


V A A

where we have used the heat balance for each point of the surface (Figure 12). Once again, since
T is uniform, we obtain:

∫ ∇ ⋅ qdV = h (T − T∞ )A (140)
V

Substituting equations (137), (138) and (140) into equation (136) leads to

dT
ρC p V = −h (T − T∞ )A + E r V (141)
dt

Now we use the change of variable

Φ = T − T∞ (142)

to express equation (141) as follows

dΦ hA E V
=− (Φ − r ) (143)
dt ρC p V hA

This equation can be separated and integrated from t=0, Φ= Φ0 (Φ0=T0-T∞) to time t to get, after
manipulations,
E V E V ⎛ hA ⎞⎟
Φ = r + (Φ 0 − r ) exp⎜ − t (144)
hA hA ⎜ ρC p V ⎟
⎝ ⎠

When t→∞, a steady state is reached in which

E V
Φ ss = r (145)
hA
32

Note that equation (144) can be expressed in terms of this steady state value as

⎛ hA ⎞⎟
Φ = Φ ss + (Φ 0 − Φ ss ) exp⎜ − t (146)
⎜ ρC p V ⎟
⎝ ⎠

The steady state value of the temperature can also be found by a steady state energy balance:

⎧Energy per unit time lost ⎫


⎪ ⎪ ⎧Thermal energy per unit time ⎫
⎨from the body to the ⎬=⎨ ⎬ (147)
⎪surroundings by convection ⎪ ⎩generated inside the body ⎭
⎩ ⎭
Or:
hA (Tss − T∞ ) = E r V (148)

which leads directly to equation (145). A special case occurs when Er=0, which implies that the
steady state will have T=T∞, and equation (146) simplifies to

⎛ hA ⎞⎟
Φ = Φ 0 exp⎜ − t (149)
⎜ ρC p V ⎟
⎝ ⎠

Figure 13 shows the trends followed by equation (146).


The heat transfer rate from the body to the surroundings will be a function of time also. We
can evaluate the instantaneous heat transfer rate by

Q = hA (T − T∞ ) = hAΦ (150)

which implies, using equation (144),

⎛ hA ⎞⎟
Q = E r V + (hAΦ 0 − E r V) exp⎜ − t (151)
⎜ ρC p V ⎟
⎝ ⎠

Note that the steady state value is Q=ErV. The total amount of thermal energy that leaves the
body between t=0 and a specified time t is

~ t
Q = ∫ Qdt (152)
0

Substituting equation (151) and integrating leads to

~ E V ⎡ ⎛ hA ⎞⎟⎤
Q = E r Vt + ρC p V(Φ 0 − r ) ⎢1 − exp⎜ −
⎜ ρC p V ⎟⎥⎥
t (153)
hA ⎢⎣ ⎝ ⎠⎦
33

For the special case Er=0, this equation simplifies to

~ ⎡ ⎛ hA ⎞⎟⎤
Q = ρC p VΦ 0 ⎢1 − exp⎜ −
⎜ ρC p V ⎟⎥⎥
t (154)
⎢⎣ ⎝ ⎠⎦

Φ Φ ss>Φ 0

Φ0

Φ ss<Φ 0

Φ ss=0

Figure 13. Evolution of the temperature of a body with negligible internal resistance.

Note that ρCpVΦ0 represents the maximum amount of thermal energy that can leave the body
~
(and it is, therefore, the value of Q as t→∞).

Example: heat transfer from a thermocouple junction


A thermocouple is a device used to measure temperature based on the Seebeck effect (also
called the thermoelectric effect): when a metal is subjected to a temperature gradient, a gradient
in electric field potential is generated within it. Thus, if a wire joins two points in space that are at
different temperatures, a voltage drop is generated between the points. Thermocouples work by
joining together two wires made of dissimilar metals and measuring the difference in the voltage
drops between them when they are joined at two junctions that are at different temperatures. The
measured difference is correlated to the temperature difference between the junctions by a
calibration curve. Since thermocouples are relatively cheap and accurate, they are the most
widely used method to measure temperatures. An important aspect to be considered when using
thermocouples (especially when measuring temperatures that change with time) is the response
time of the thermocouple. Consider the measurement of the temperature of a fluid (T∞) using a
thermocouple. The junction (Figure 14) will be a solder bead used to join the wires of the two
materials. Let us consider that the solder bead is a sphere of diameter D=0.71 mm, the thermal
conductivity of the solder material is k=20 W/m K, while its density and specific heat are ρ=8500
34

kg/m3 and Cp=400 J/kg K.

Figure 14. A thermocouple junction consists of two wires of dissimilar materials joined by a
solder bead. In this case, the objective is to measure the temperature of the flowing fluid, T∞.

In addition, T∞=200ºC, and the bead is initially at T0=25ºC, just before it is immersed in the
fluid. The convective heat transfer coefficient between the bead and the moving fluid is h=400
W/m2K. We would like to calculate how long it will take the thermocouple to record a
temperature that is close enough to the fluid's temperature. We will specify an accuracy of 1ºC in
the measurement.
First, we calculate the Biot number for this process:

h (D / 2)
Bi = = 0.0071
k

Since Bi«1, we can consider that internal heat transfer resistance in the bead is negligible.
Therefore, the temperature in the bead is uniform. To get a measurement within 1ºC of 200ºC, we
will find the time necessary for the temperature of the bead to be T=199ºC.
The temperature of the bead as a function of time is given by equation (149). Solving that
equation for time yields
ρC p V ⎛ T − T∞ ⎞
t=− ln⎜⎜ ⎟⎟ = 5.2 s
hA ⎝ T0 − T∞ ⎠

This means that we would have to wait 5.2 s to record the fluid temperature with the desired
accuracy.

Transient heat conduction with important internal resistance


In general, the transient temperature inside a body will depend on position. In this case, the
heat conduction equation becomes a partial differential equation whose solution will yield the
temperature of the body as a function of position and time. For bodies of complex shape, the
solution needs to be found numerically. There are a few shapes for which analytical solutions are
available. Here we will consider some of these shapes, starting with a flat wall.
Consider a flat wall of thickness 2L (Figure 15) initially at T=T0 that will exchange thermal
energy with a fluid at T=T∞. We will consider that Er=0, and that the wall has a constant thermal
conductivity. It is our objective to find T=T(t,x). Under the conditions considered, equation (131)
simplifies to

∂T ∂ 2T
=α (155)
∂t ∂x 2
35

Figure 15. Transient heat transfer from a flat wall. The wall is very large in the y and z
directions. It is initially at a uniform temperature T0 and it is brought in contact at t=0 with a fluid
at T∞.

The initial condition is


T=T0, t=0 (156)

We will solve the problem in half of the solution domain (0≤x≤L) taking advantage of the
geometric symmetry. With this in mind, we can impose the following symmetry condition as a
boundary condition,
∂T
= 0 , x=0 (157)
∂x

while the second boundary condition will be a heat balance on the surface exposed to the fluid,

∂T
−k = h (T − T∞ ) , x=L (158)
∂x

The solution T=T(t,x) depends on all the parameters present in the differential equation and
boundary conditions (α, T0, k, h, T∞, L). We will use dimensionless variables to minimize the
number of parameters in the formulation. Consider the definition of the following dimensionless
variables:
T − T∞
Θ= (159)
T0 − T∞

x
η= (160)
L

If we solve these two equations for T and x, and substitute the result in the differential equation
(155), we get
36

∂Θ α(T0 − T∞ ) ∂ 2 Θ
(T0 − T∞ ) = (161)
∂t L2 ∂η2

We note that all parameters are eliminated from the differential equation by defining a
dimensionless time such that

αt
τ= (162)
L2

The differential equation becomes

∂Θ ∂ 2 Θ
= (163)
∂τ ∂η2

The initial (equation 156) and boundary (equations 157 and 158) conditions become

Θ=1, τ=0 (164)

∂Θ
= 0 , η=0 (165)
∂η

∂Θ
= −BiΘ , η=1 (166)
∂η
where
hL
Bi = (167)
k

The dimensionless formulation (equations 163 to 166) has only one parameter (Bi). This problem
has an analytical solution that can be found by separation of variables. To do this, we start by
postulating a solution of the form

Θ(τ, η) = f (η)g(τ) (168)

After substitution in the differential equation and separation, we obtain

1 dg 1 d 2 f
= = −λ2 (169)
g dτ f dη2

The left-hand side of this equation only depends on τ while the right-hand side only depends on
η, so both must be equal to a constant (-λ2) whose sign has been chosen as negative to ensure an
exponential decay with time (the choice is really irrelevant if imaginary numbers are allowed in
the derivation). The differential equation for g can be directly integrated:
37

dg
= −λ2 dτ ⇒ g = Ae− λ2 τ (170)
g

while for f we get

d 2f
+ λ2 f = 0 (171)
dη2

whose solution is
f = C1 sin(λη) + C 2 cos(λη) (172)

The boundary conditions (165) and (166) are separable:

df df
g =0⇒ = 0 , η=0 (173)
dη dη

df df
g = −Bi fg ⇒ = −Bi f , η=1 (174)
dη dη

Application of equation (173) yields

C1λ cos(λη) − C 2 λ sin(λη) η = 0 = 0 ⇒ C1=0 (175)

From equation (174) we find

− C 2 λ sin λ = −C 2 Bi cos λ (176)

Since C2 cannot be zero (this would lead to the trivial solution), we conclude that

λ tan λ = Bi (177)

This equation has infinite solutions (Figure 16). The infinite roots of this equation (λ1<λ2<λ3...)
are the eigenvalues of the Sturm-Liouville problem posed by equations (171), (173) and (174).
Equation (177) is the eigenvalue condition. This means that this problem has infinite solutions of
the form (172) (with C1=0), one for each eigenvalue:

f n (η) = C 2 cos(λ n η) (178)

Similarly, there will be infinite functions g(τ) from equation (170)

2
g n (τ) = Ae − λ n τ (179)

The general solution of the partial differential equation is the linear combination of all the
solutions given by equation (168):
38

2.5

2 tanλ

1.5
λ1
1
Bi/λ
0.5 λ2
λ3
0

-0.5

-1
0 1 π/2 2 3 4 3π/2 5 6 7
λ
Figure 16. Graphical solution of equation (177), interpreted as the intersection between the
functions tanλ and Bi/λ. Values in the plot are for the case Bi=1. The first three roots are shown,
but the periodicity of tanλ shows that infinite solutions exist.


2
Θ(τ, η) = ∑ a n cos(λ n η)e − λ n τ (180)
n =1

The coefficients an are evaluated by using the initial condition (164):


1= ∑ a n cos(λ n η) (181)
n =1

To find each coefficient, we will use the orthogonality condition satisfied by solutions of a
Sturm-Liouville problem, which in this case states that, for any two different eigenvalues λn, λm,

1
∫ cos(λ n η) cos(λ m η)dη = 0 , for n≠m (182)
0

We now multiply both sides of equation (181) by cos(λmη) and integrate from η=0 to η=1.
Equation (182) implies that all the terms in the sum on the right-hand side of the resulting
equation will be zero, except the term for which n=m,
39

1 1
∫ cos(λ m η)dη = a m ∫ cos 2 (λ m η)dη (183)
0 0

Evaluating the integrals leads to

2 sin λ m
am = (184)
λ m + sin λ m cos λ m

Letting m=n and substituting into equation (180) we find the final solution:

∞ 2 sin λ 2
Θ(τ, η) = ∑ λ n + sin λ n ncos λ n cos(λ n η)e − λ n τ (185)
n =1

The Fourier series in equation (185) is uniformly convergent. The factor that controls its
convergence is the exponential term, which rapidly becomes small as n increases (recall that λn
increases with n). Note that, the larger the value of τ, the faster the decay of the exponential with
increasing n. In fact, detailed calculations show that, if Bi≥0.01, the first term of the series is
enough to give the value of Θ with accuracy of 0.1% or better, as long as τ≥0.1. For lower Biot
numbers, it may be safe to assume that internal resistance to heat transfer is negligible and the
process follows the simpler analysis presented in the previous section. For this case, the
approximate solution is

2 sin λ1 2
Θ(τ, η) ≅ cos(λ1η)e − λ1 τ , Bi≥0.01, τ≥0.1 (186)
λ1 + sin λ1 cos λ1

Values of λ1 (first root of the eigenvalue condition 177) are shown in Figure 17 as a function of
the Biot number. Note that λ1→0 as Bi→0 (not in the plot because of the semi-log scale), and
that λ1→π/2 (1.5708) as Bi→∞ (equation 177).
Now that the temperature profile is known, the heat transfer rate exchanged between the wall
and the surroundings can be evaluated by calculating the heat flux at the surface (qx) as either
side of equation (158). A different way to do it is to express the heat flux in terms of the volume-
averaged temperature, defined by

1
V∫
< T >= TdV (187)
V

where V is the volume of the body (the wall in this case). Let us return to the heat conduction
equation (135) (with Er=0). We take the volume average of this equation as follows,

1 ∂T 1
V ∫ ρC p dV = − ∫ ∇ ⋅ qdV
∂t V
(188)
V V

The averaging operator can be inserted into the time derivative, and the right-hand side can be
40

expressed as an area integral using the divergence theorem to get

1.6
λ1
1.4

1.2

0.8

0.6

0.4

0.2

0
0.01 0.1 1 10 100
Bi
Figure 17. First root of equation (177) as a function of the Biot number.

∂ ⎛⎜ 1 ⎞
⎟ = − 1 q dA
ρC p

∂t V ∫ TdV
⎟ V∫
n (189)
⎝ V ⎠ A

where qn is the component of the heat flux vector normal to the surface of the body, A. The
integration of qn over A yields the total heat transfer rate from the body to the surroundings, Q.
Therefore, equation (189) can be written as

d<T>
Q = −ρC p V (190)
dt

Note that, since <T> is only a function of time, we have turned the partial time derivative into a
total derivative.
Equation (190) is a macroscopic energy balance, representing the conservation of energy
principle applied to the whole volume V. It applies to any control volume V, as long as Er=0, and
ρ and Cp are constant and uniform. An interesting consequence of this equation is that, regardless
of the temperature profile inside the body, the average temperature's change with time is
controlled by the heat transfer rate with the surroundings. For example, if Q=0 (insulated body),
equation (190) states that <T> is a constant, even though the temperature profile inside the body
may be changing with time.
Going back to the solution for the flat wall, a dimensionless average temperature can be
defined by an equation similar to (187). In this case we can take V=2AwL, where Aw is the
surface area of one side of the wall, and, since the temperature dependence on position is only on
x (or η), dV=Awdx, so that
41

1 L
2L ∫
< Θ >= Θdx (191)
−L

Using equation (160) and taking advantage of the symmetry of Θ around η=0 yields

1
< Θ >= ∫ Θdη (192)
0

Substituting the solution (equation 185) into this equation and integrating leads to

∞ 2 sin 2 λ 2
< Θ >= ∑ (λ n + sin λ n cosn λ n )λ n e − λ n τ (193)
n =1

From the definition of dimensionless temperature (equation 159) we can show that

< T >= T∞ + (T0 − T∞ ) < Θ > (194)

Using also equation (162), we find the average temperature as a function of time:

∞ 2 sin 2 λ n ⎛ λ2 αt ⎞
< T >= T∞ + (T0 − T∞ ) ∑ exp⎜⎜ − n ⎟⎟ (195)
(λ + sin λ n cos λ n )λ n
n =1 n ⎝ L ⎠
2

The instantaneous heat transfer rate can be calculated from equation (190):

4k ∞ λ n sin 2 λ n ⎛ λ2 αt ⎞
Q= A w (T0 − T∞ ) ∑ exp⎜⎜ − n ⎟⎟ (196)
L λ + sin λ n cos λ n
n =1 n ⎝ L ⎠
2

Note that this represents the thermal energy per unit time that leaves the wall through both
surfaces (x=±L). The total amount of heat transferred from t=0 to t can be calculated by
integration of Q (as stated in equation 152) or, alternatively, direct integration of equation (190),
which results in
~
Q = ρC p V(T0 − < T >) (197)

Substitution of equation (195) into this equation yields

~ ⎡ ∞ 2 sin 2 λ n ⎛ λ2 αt ⎞⎤
Q = ρC p V (T0 − T∞ ) ⎢1 − ∑ exp⎜⎜ − n ⎟⎟⎥ (198)
⎢⎣ n =1(λ n + sin λ n cos λ n )λ n
2
⎝ L ⎠⎥⎦

The analysis developed in this section can be applied to heat transfer from bodies with other
simple geometries. Analytical solutions can be found by using separation of variables for
transient heat conduction from an infinitely long cylinder and from a sphere. These are presented
42

without details in what follows:

(1) Infinitely long cylinder (radius R)


Here we consider that T=T(t,r). The dimensionless variables used are

T − T∞
Θ= (199)
T0 − T∞

αt
τ= (200)
R2

r
ξ= (201)
R

hR
Bi = (202)
k

The dimensionless conduction equation simplifies to

∂Θ 1 ∂ ⎛ ∂Θ ⎞
= ⎜ξ ⎟ (203)
∂τ ξ ∂ξ ⎜⎝ ∂ξ ⎟⎠

and the initial and boundary conditions are

Θ=1, τ=0 (204)

∂Θ
= 0 , ξ=0 (205)
∂ξ

∂Θ
= −BiΘ , ξ=1 (206)
∂ξ

In this case, the solution is expressed in terms of Bessel functions. The eigenvalue condition is

J (λ )
λ n 1 n = Bi (207)
J 0 (λ n )

The first root of this equation is presented in Figure 18. The solution is

2
∞ 2 J1 (λ n ) −λ τ
Θ= ∑ J 0 ( λ n ξ) e n (208)
n =1 λ n J 0 (λ n ) + J1 (λ n )
2 2

Once again, the first term of the series is a good approximation for Bi≥0.01 and τ≥0.1. The first
43

coefficient of the series is

2 J1 (λ1 )
a1 = (209)
λ1 J 0 (λ1 ) + J12 (λ1 )
2

and is shown as a function of Biot number in Figure 19.

λ1 sphere
3

2.5 cylinder
2

1.5 flat plate

0.5

0
0.01 0.1 1 10 100
Bi
Figure 18. First eigenvalue as a function of the Biot number for flat plate (same as in Figure 17),
infinitely-long cylinder (first root of equation 207), and sphere (first root of equation 215)

1.7
a1
1.6

1.5

1.4

1.3

1.2

1.1

1
0.01 0.1 1 10 100
Bi
Figure 19. First coefficient of the Fourier series for the cylinder, equation (209).
44

The total heat transferred is given by

~ 2
Q ∞ 4 J12 (λ n ) − λnτ
= 1− ∑ e (210)
ρc p V(T0 − T∞ ) 2 2 2
n =1 λ n J 0 (λ n ) + J1 (λ n )

(2) Sphere (radius R)


Here we consider T-T(r,t). The dimensionless variables are defined in the same way as for the
cylinder (equations 199 to 202), but using the radius of the sphere. The heat conduction equation
is
∂Θ 1 ∂ ⎛ 2 ∂Θ ⎞
= ⎜ξ ⎟ (211)
∂τ ξ 2 ∂ξ ⎜⎝ ∂ξ ⎟⎠

and the initial and boundary conditions are

Θ=1, τ=0 (212)

∂Θ
= 0 , ξ=0 (213)
∂ξ

∂Θ
= −BiΘ , ξ=1 (214)
∂ξ

The eigenvalue condition is

1 − λ n cot λ n = Bi (215)

The first eigenvalue is plotted in Figure 18. The solution is

2
∞ 4 sin λ n − λ n cos λ n sin(λ n ξ) − λ n τ
Θ= ∑ e (216)
n =1 λ n 2λ n − sin 2λ n ξ

Total heat transferred:

~ 2
∞ 12 (sin λ − λ cos λ ) 2 − λ τ
Q
= 1− ∑ n n n e n (217)
ρc p V(T0 − T∞ ) 3
n =1 λ n 2λ n − sin 2λ n

You might also like